InformAction Forums

Steve Gibson wrote:STEVE: The bad news is that just three days ago,[[5 Oct -- TT.]] this last Monday, a fake PayPal certificate was posted on the Internet which allows SSL connections, Secure Socket Layer connections, to be spoofed, that is, like with various sorts of phishing attacks, using this fraudulent certificate.

ALEX: So you can think you're paying through PayPal, and you're really paying - you're really connecting to somebody else.

STEVE: Well, exactly. And in fact this is - what's disturbing about this is that this is more than two months old. This is a defect that originally affected all Windows browsers. Microsoft still to this day, more than nine weeks after this went public, has not fixed this. The problem still exists in their own CryptoAPI library, which is a shared component of Windows that IE, Apple Safari running on Windows, and Chrome all use. So today IE, Safari on Windows, and Chrome are all vulnerable to this.

What's interesting is that Firefox, both version chains of Firefox, the 3.0 and the 3.5, fixed it within a couple days. And there was also the problem, even in the Mac OS X originally, but Apple fixed it a couple weeks later. So this has long been fixed for Firefox under Windows and Safari under OS X. But even now, more than two months later, has not been fixed for IE, Apple Safari under Windows, and Chrome. And now we have this fake PayPal certificate that is being circulated on the Internet that essentially, if it's used, you can actually establish an SSL connection to what looks like PayPal. We talked about this a couple months ago, to remind our listeners.

The idea is that the way strings are stored in pretty much all modern operating systems is it's just a sequence of bytes that ends with a null byte, that is, a zero byte. They're so-called null terminated strings. Strings historically have been stored in various different formats. Pascal was famous for storing the length of the string as the first byte, which was convenient for all kinds of reasons. The problem was that a byte can only be 0 to 255. So Pascal strings could never be longer than 255 characters. So that was sort of fixed by saying, wait a minute, we'll just allow a string to be any length, but terminate it with a null. Well, of course that has had disastrous security consequences. That whole null terminated string issue, while it's very convenient for programmers to sort of scan until you hit a null, that's largely responsible for all the buffer overrun problems we have today.

ALEX: Now, is part of that giving a hacker an idea of what to look for?

STEVE: Well, it's more that the operating system will just read bytes until it hits zero. And it'll do that blindly. So if you tell it to, like, copy one string to somewhere else, it will copy as much as you give it until it hits a null character. So it creates this, like, all kinds of opportunities for exploitation. What's interesting about this particular null - it's call the "null prefix vulnerability" in certificates - is that you can create a certificate, http://www.paypal.com null, that is, a zero, and then anything else you want, like mymachine.secure.net. And so the certificate is actually for the secure.net domain. And certificate issuers will issue certificates to the secure.net domain. And then if you embed a null between sort of your own machine name, http://www.paypal.com, the browsers, while they're parsing the name on the certificate, they stop at the first null. Which is the way strings are processed in our modern operating systems. So it's not a surprise that they do this. But you absolutely don't want that behavior in this particular instance.

ALEX: Right.

STEVE: So anyway, so essentially this is a bad problem. Microsoft has not responded. And as of three days ago there is this known fraudulent PayPal very spoofable certificate floating around the 'Net. And of course the big question is, okay, we know about the PayPal cert. What other ones have been issued that we don't know about?

ALEX: Right.

STEVE: So essentially the takeaway from this is, at the moment you cannot trust, that is, a Windows user cannot trust any Windows browser other than Firefox. The Firefox guys fixed this. They took responsibility away from the underlying Windows platform and fixed it themselves within days.

ALEX: So, now, so Apple could fix it, for instance, on Safari. They just - they would need to - but right now they're relying too much on the Windows framework?

STEVE: Exactly. They're using - there's a shared library called the CryptoAPI that IE, Safari on Windows, and Chrome all just assume the underlying framework is reliable. They're all using it. And as a consequence, today they're all vulnerable. And so Apple did fix it in OS X immediately, but haven't done so on Safari under Windows, probably presuming that this is, you know, hey, this is not our fault. This is a Windows problem. The problem is that it makes all browsers except Firefox untrustworthy, I mean, completely untrustworthy, for making secure connections until Microsoft finally fixes it. Which maybe they'll do so soon. I hope so.

GµårÐïåñ wrote:I have a credit provider whose backed by GE Money and they use the following website: Account Management

I went there today and I got a notice from both Perspective that the certificate is invalid showing signs of being attacked and SSL Blacklist also said it was invalid and additionally Trusteer's Rapport also gave a warning that the certificate is invalid and should be avoided.